Pharm Res (2018) 35:151 https://doi.org/10.1007/s11095-018-2432-3 RESEARCH PAPER Development of a Simple Mechanical Screening Method for Predicting the Feedability of a Pharmaceutical FDM 3D Printing Filament 1 2 1,3 4 1 Jehad M. Nasereddin & Nikolaus Wellner & Muqdad Alhijjaj & Peter Belton & Sheng Qi Received: 28 March 2018 /Accepted: 17 May 2018 # The Author(s) 2018 Conclusion The screening method developed in this study ABSTRACT Purpose The filament-based feeding mechanism employed showed, with statistical significance and reproducibility, the by the majority of fused deposition modelling (FDM) 3D ability to predetermine the feedability of extruded filaments printers dictates that the materials must have very specific into an FDM printer. mechanical characteristics. Without a suitable mechanical KEY WORDS feedability screening fused deposition profile, the filament can cause blockages in the printer. The . . . purpose of this study was to develop a method to screen the modeling 3D printing hot melt extrusion plasticization mechanical properties of pharmaceutically-relevant, hot-melt printability solid dispersions extruded filaments to predetermine their suitability for FDM. Methods A texture analyzer was used to simulate the forces a filament is subjected to inside the printer. The texture analyzer ABBREVIATIONS produced a force-distance curve referred to as the flexibility pro- 3DP 3D printing file. Principal Component Analysis and Correlation Analysis sta- ABS Acrylonitrile butadiene styrene tistical methods were then used to compare the flexibility profiles ATR-FTIR Attenuated Total Reflectance Fourier Transform of commercial filaments to in-house made filaments. Infrared Results Principal component analysis showed clearly separated CPP Critical processing parameters clustering of filaments that suffer from mechanical defects versus DSC Differential Scanning Calorimetry filaments which are suitable for printing. Correlation scores FDA Food and Drug Administration likewise showed significantly greater values with feedable fila- FDM Fused Deposition Modeling ments than their mechanically deficient counterparts. HME Hot-Melt Extrusion HPC hydroxypropyl cellulose HPMCAS Hypromellose acetate succinate Guest Editor: Dennis Douroumis PAC Paracetamol Electronic supplementary material The online version of this article PCA Principal Component Analysis (https://doi.org/10.1007/s11095-018-2432-3) contains supplementary PEG Polyethylene glycol material, which is available to authorized users. PEO Polyethylene glycol PLA Polylactic acid * Sheng Qi PVA Polyvinyl alcohol email@example.com PXRD Powder X-Ray Diffraction TA Texture Analysis School of Pharmacy, University of East Anglia, Norwich, Norfolk NR4 7TJ, UK Norwich Research Park, Quadram Institute Bioscience, Colney Norwich, Norfolk NR4 7UA, UK INTRODUCTION Department of Pharmaceutics, College of Pharmacy, University of Basrah, Basrah, Iraq In recent years, there has been a rise in interest in utilizing 3D printing (3DP) as means to manufacture pharmaceutical dos- School of Chemistry, University of East Anglia, Norwich, Norfolk NR4 7TJ, UK age forms due to its ability to produce bespoke objects 151 Page 2 of 13 Nasereddin et al. (2018) 35:151 possessing high geometrical complexity quickly with high pre- Machine-specific parameters are parameters relating to the make cision and accuracy. This gives 3DP the potential for the and model of the printer, operation-specific parameters are pa- manufacturing of personalized dosage forms (1,2). FDM is a rameters relating to the processing conditions during the printing variant of 3DP that utilizes filament-shaped thermoplastic run, and material-specific parameters are parameters relating to polymers as the building material. The mechanical assembly the physiochemical properties of the material being printed. of a FDM printing head consists of the feeding rollers, heating Most studies reported in the literature utilize commercially avail- zone, and the nozzle (3–5). The filament is fed into the printer able, hobbyist application printers which do not allow for control by the action of the two counter-rotating feeding rollers, as of the machine-specific parameters. While such printers allow illustrated in Fig. 1a. The filament is fed to a heating zone control over operation-specific parameters such as printing speed where the filament is melted and extruded through the print- and temperature, restrictions imposed by polymer melt rheology ing nozzle onto the build plate layer-wise to form the desired and the printer feeding mechanism generally result most phar- object. maceutical polymers being unsuitable for FDM applications. For pharmaceutical applications of FDM, the incorpora- Although the CPPs of FDM have been identified and are rela- tion of a drug substance into the filament is achieved by two tively well understood, there exists, to the best of our knowledge, methods, impregnation and hot melt extrusion (HME) (2). no preformulation tools have been developed to allow for a quick Impregnation is done by immersing the filament in an organic and rational design of pharmaceutical FDM formulations. Most solution of the drug. This method often yields low levels of formulations reported in the literature employed a trial-and- drug loading (6–8). Preparation of FDM filaments by HME is error approach for development. the more attractive method for pharmaceutical applications as To achieve a continuous, high-throughput FDM printing it allows for higher drug loading than the impregnation meth- operation, the filament needs to be firstly feedable through od, and allows for the use of pharmaceutical-grade polymers the feeding zone of the printer, and must also possess suitable (9–11). Most hot melt extrudable pharmaceutical grade poly- melt flow properties to be printable once is transferred into the mers, however, do not possess the required properties to allow heating zone of the printing head. The filament-based feeding for good quality FDM printing. This has been recognized as a mechanism employed by FDM printers utilizes mechanical significant technical barrier for further developing the phar- gearing arrangements to push the filament into the heating maceutical applications of FDM printing (10). zone (Fig. 1a) (4). In order to accurately control the feeding rate, Critical process parameters (CPPs) which govern FDM are the filament has to be held tightly (pinched) between the two categorised into three types of parameters: machine-specific, rollers (13), leaving it effectively under compression throughout operation-specific, and material-specific parameters (12). the feeding process. In situations where the filament is brittle, this is likely to cause the filament to fracture, discontinuing the forwardpropulsionofthe filament,causing ablockageinthe printing head, as illustrated in Fig. 1a. Blockages in the printing head are very problematic; broken pieces of filaments inside the printing head can contaminate the machine, compromising the purity of any dosage forms one wishes to fabricate. Blockages in the printing head are also difficult to clean, often requiring disassembly of theentireprintingheadtobecleared. Therefore, there is a need for screening the mechanical suitabil- ity of the filament before attempting to feed the filament into a printer. For the scope of this article, the term feedability is used to de- scribe the mechanical suitability of a filament for FDM, with non-feedable filaments being filaments that can cause a block in the printing head of an FDM printer, regardless of whether their melt flow rheology is within acceptable limits for FDM. This study describes the development of a new formulation screening tool for the predetermination of the feedability of FDM filaments fabricated by HME from pharmaceutically rel- evant polymers. This tool was built based on the understanding of the relationship between the mechanical properties of some pharmaceutical polymer blends and how they correlate with their suitability for FDM. A number of simple in-house filaments, Fig. 1 (a) Illustration of the different behaviour of filaments during feeding; (b) the texture analysis filament feedability test rig. commercial filaments, as well as a printable complex Development of a Simple Mechanical Screening Method (2018) 35:151 Page 3 of 13 151 pharmaceutical filament previously reported in the literature (the 4000) and paracetamol (PAC) were purchased from Sigma material was a placebo filament containing Eudragit EPO, Aldrich (Sigma Aldrich, Salisbury, United Kingdom). Tween 80, PEG and PEO named as EUD was used in this study) (10) were prepared and attempted to be fed into an unmodified, commercial 3D printer to determine their feedability. A custom- Preparation of In-House Filaments made texture analyzer rig was used to test the filament response in a compress-and-release cycle, yielding a plot of force (exerted In-house filaments were prepared by HME, using a Haake by the filament as resistance to deformation) vs.distancecom- Minilab II hot melt compounder (Thermo Fisher Scientific, pressed plot. The texture analysis experiments were used to Karlsruhe, Germany) equipped with a 1.75mm circular die. A quantify the mechanical properties of the filaments and to ascer- list of prepared formulations and their key extrusion parameters tain whether it is possible to predetermine the feedability of a can be found in Table I. All multi-component formulations were filament without having to compromise the printing head. cycled in the extruder for 5 min at a screw speed of 100 RPM to Using the force/distance plots (hereinafter referred to as the ensure homogenous mixing (10). Following extrusion, filaments flexibility profile) produced by commercial filaments as a control, with diameters of 1.75mm ± 0.05 mm were collected for further correlation analysis and principle component analysis (PCA) testing. were used to determine whether there exists a statistically signif- icant correlation between the flexibility profiles of different fila- ments and their feedability (and subsequently printability). This Filament Characterization allows one to predetermine if hot melt extruded filament pos- sesses adequate mechanical properties to be feedable. A test Differential Scanning Calorimetry (DSC) which can be used as a performulation tool to minimize trial- and-error when developing pharmaceutical formulations for DSC was conducted using a Q20 differential scanning calo- FDM. The relationship between the formulation composition rimeter (TA Instruments, Newcastle, United States). All in- and the feedability of the filaments investigated in this study house prepared filaments were tested using a heat-cool- can bring new insights into the development of principles in reheat cycle with a temperature range of 20°C to 185°C at rationalization of FDM formulation design. 10°C/min. All samples were tested as fresh samples immedi- ately after extrusion. All tests were done in triplicates. MATERIALS AND METHODS Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR) Materials FTIR was conducted using a Vertex 70 FTIR spectrometer Acrylonitrile butadiene styrene (ABS) and Makerbot (Bruker Optics Ltd., United Kingdom), equipped with a Dissolvable Filament commercial filaments were purchased from MIRacle™ single reflection ATR accessory (Pike Makerbot (Makerbot Industries LLC., New York, United Technologies, United States) fitted with a diamond internal States). Polylactic acid (PLA) commercial 3D printing filaments reflection element. ATR-FTIR spectra were acquired in ab- −1 were purchased from XYZprinting (XYZprinting Inc., sorbance mode, using a resolution of 2 cm , 32 scans for each California, United States). All three commercial filaments were sample, within the range of wavenumbers from 4000 to −1 used as purchased. Pre-plasticized polyvinyl alcohol (other- 550 cm . Spectra analysis was conducted using OPUS ver- wise known as Mowiflex ) C-17 grade pellets were graciously sion 7.8 (Bruker Optics Ltd., United Kingdom). All measure- donated by Kurary (Kurary GmbH, Frankfurt, Germany). ments were done in triplicate. Hypromellose acetate succinate (HPMCAS, low-fine grade) was graciously donated by Shin Etsu (Shin Etsu Inc., Tokyo, Japan). Polysorbate (Tween 80) was purchased from Acros Powder X-Ray Diffraction (PXRD) Organics (Acros Organics, Geel, Belgium). Polyethylene oxide N-10 grade (PEO; molecular weight = 100,000) was graciously A Thermo ARL Xtra X-ray diffractometer (Thermo donated by Colorcon (Colorcon Ltd., Dartford, United Scientific, Switzerland) equipped with a copper X-ray Tube Kingdom). Eudragit® EPO was graciously donated by Evonik (k = 1.540562 Å) was used to detect the presence of drug crys- industries (Evonik, Darmstadt,Germany). Soluplus and tals (if any) in the extruded filaments. A scanning range of 3° Kollidon vinyl acetate 64 (polyvinyl pyrrolidone vinyl acetate <2θ < 30°, using a step scan mode with step width of 0.01° 64) were graciously donated by BASF (BASF inc., Ludwigshafen, and scan speed of 1 s/step was used to conduct all Germany). Polyethylene glycol 4000 (PEG; molecular weight = measurements. 151 Page 4 of 13 Nasereddin et al. (2018) 35:151 Table I Compositions and Formulation Constituents Extrusion temperature Extrusion Conditions of the Filaments Mowiflex® Mowiflex® (contains PVA and an undeclared plasticizer at an 170°C undeclared concentration HPMCAS HPMCAS (100% w/w) 170°C PEO PEO (100% w/w) 75 °C PVP/VA 64 PVP/VA 64 (100% w/w) 140°C Soluplus® Soluplus® (100% w/w) 120°C Eudragit® EPO Eudragit® EPO (100% w/w) 120°C EUD Eudragit EPO (55.5% w/w) + 11.1% w/w, 16.7% w/w, and 16.7% 100 °C Tween 80, PEG, and PEO, respectively HD HPMCAS (90% w/w) and paracetamol (10% w/w) 150°C HP10 HPMCAS (90% w/w) and PEO (10% w/w) 150°C HP10D HPMCAS (81% w/w), PEO (9% w/w) and paracetamol (10% w/w) 140°C HP20 HPMCAS (80% w/w) and PEO (20% w/w) 150°C HP20D HPMCAS (72% w/w), PEO (18% w/w) and paracetamol (10% w/w) 130°C HP30 HPMCAS (70% w/w) and PEO (30% w/w) 140°C HP30D HPMCAS (63% w/w), PEO (27% w/w) and paracetamol (10% w/w) 120°C HP40 HPMCAS (60% w/w) and PEO (40% w/w) 130°C HP70 HPMCAS (30% w/w) and PEO (70% w/w) 100 °C HP90 HPMCAS (10% w/w) and PEO (90% w/w) 85°C SP Soluplus® (90% w/w) and PEG (10% w/w) 110°C ST Soluplus® (80% w/w) and Tween® 80 (20% w/w) 100 °C FDM Feedability Testing TA Data Manipulation and Statistical Analysis Feedability of the extruded filaments was tested by feeding into a Since the materials tested were of varying hardness, the scaling of standard Makerbot® Replicator 2X commercial FDM 3D theflexibility profile(force(N) vs. distance (mm)) plots was not printer (Makerbot Industries LLC., New York, United States). directly comparable without data range normalization. Successful extrusion of the polymer through the nozzle tip was Therefore, data range normalization was performed using the regarded as successful feeding, making the filament feedable.It equation should be highlighted that the printing quality was not assessed and is out of the scope of this study. All filaments were fed at the Y ¼ ð1Þ Normalized ΣY printer’s default printing temperature of 230°C. n th where Y is the n point on the Y-axis (force). Correlation anal- ysis of the flexibility profiles of the pharmaceutical filaments with Texture Analysis (TA) Screening Test those of the commercial filaments was conducted using Microsoft® Excel 2016 expanded with the Data Analysis add- Compression tests that simulate the feeding process of the fila- on. PCA was conducted using IBM® SPSS Statistical Analysis ment through the printing head were performed using a Suite (version 25), with the Varimax rotation method, 25 itera- TA.XT2 Plus Texture Analyzer (Stable Micro Systems, tions for convergence, extracting components possessing an ei- Godalming, United Kingdom) equipped with an in-house rig genvalue ≥1. The loadings of the three components with an (Fig. 1b) and a 5 kg Load cell. Filaments were compressed axially eigenvalue at or above 1 are reported in Table S1 in the with a compression speed of 3.15 mm/sec, corresponding to the Supplementary Materials. roller movement speed of a Makerbot® Replicator 2X (deter- mined by feeding an accurately cut 10 cm filament into the printer head and measuring the time needed for the filament to pass through the printing head). 5 cm long filament pieces RESULTS were held standing in conical end caps to allow bending and to avoid fracturing them with the clamps. The compression dis- Filament Characterization tance was set to 15 mm with a trigger force of 0.05 N and data was collected during both compression and release. TA tests were There are 3 categories of pharmaceutical filaments produced done in triplicate for all tested filaments. and tested in this study; pure pharmaceutical polymers which Development of a Simple Mechanical Screening Method (2018) 35:151 Page 5 of 13 151 Table II Measured and are hot melt extrudable (HPMCAS, PEO, PVPVA 64, Soluplus, Formulation T (°C) Predicted T ’sof the g Eudragit EPO), polymer-plasticizer blends (HPMCAS-PEO, Filaments EUD, Soluplus-PEG, Soluplus-Tween), and drug loaded fila- HPMCAS 124 ments (HPMCAS-PAC, HPMCAS-PEO-PAC). The rationale PEO −60 of including these different categories of the pharmaceutical fila- PVP/VA 64 108 ments is to test a broad range of pharmaceutical materials to Soluplus® 70 validate the screening method and develop understandings of Eudragit® EPO 45 the effects of additives and drugs on the feedability and printabil- HD 83 ity of the pharmaceutical polymers. Figure 2 shows the DSC SP 42 thermograms of the physical mixes and extruded filaments of *The T of the rest of the in-house filaments HPMCAS based filaments. The DSC results of the rest of the g were not clearly identified using DSC filaments are available in the Supplementary Materials Fig. S1 and the T of the filaments that were detectable by DSC are summarized in Table II. Melting endotherms corresponding to HPMCAS mixed well with PEO after extrusion and the signif- the T of PEO at ~60°C and T of HPMCAS at ~120°C were icant amount of HPMCAS was sufficient to prevent PEO from m g seen in all physical mixes of the placebo and drug loaded recrystallization. When increasing the PEO content to above HPMCAS-PEO blends. For drug loaded physical mixes, a small 30%, clear phase separation of crystallised PEO and HPMCAS-PEO phase can be identified. Using the melting en- melting endotherm at ~169°C can be seen corresponding to the melting of PAC form I (monoclinic form) (14). For placebo HP thalpy values of the PEO melting in the HP filaments and the enthalpy value of the pure PEO (obtained from the DSC results filaments, the melting of PEO is absent in the low-loading for- mulations (HP10, HP20, and HP30), but is clearly seen in for- of pure PEO), it is possible to estimate the degree of crystallinity mulations with PEO content above 30% w/w (HP40, HP70 and of PEO in the filaments. HP90 and HP70 have 56.3 and 51.5% HP90). This result indicates that with 10–30% PEO loading, crystallinity, respectively, which is much higher than the 30.8% Fig. 2 DSC thermograms of (a) raw materials of the HP filaments, (b) placebo HP filaments (dashed lines) and physical mixes (solid lines), (c) placebo fialments with high PEO loadings; and (d) drug loaded HP filaments (dashed lines) and physical mixes (solid lines). 151 Page 6 of 13 Nasereddin et al. (2018) 35:151 for HP40. This indicates that in high PEO content filaments melting zone of the printing head. The rest of the pharmaceu- (HP70 and HP90) the continuous phase is the semi-crystalline tical filaments were successfully fed through the FDM printer. PEO. This contrasts with low PEO content filaments (HP10–30) The feedability test results are summarized in Table III. that has the HPMCAS as the continuous phase. For HP40, as the contents of HPMCAS and PEO are close, it is reasonable to TA Screening Tests expect that there is no one polymer dominates as a continuous phase, which would contribute to the significant mechanical TA tests were used to obtain the flexibility profiles of the filaments property difference observed later in the texture analysis tests. under axial compression. The areas under the curves were then For drug loaded HP30D a small melting endotherm of PEO normalized and plotted to give the comparisons of stress on the was detected indicating the semi-crystalline nature of PEO in this filament vs. distance travelled by the probe as seen in Figs. 4 and drug-loaded filament. The T of PAC was not seen in any of the 5. Using the data, it is possible to group their behaviour to ex- drug-loaded filaments, suggesting no crystalline PAC in the fila- amine the correlation between the flexibility profile and the ments. The T values of the HP placebo and drug-loaded fila- feedability of the filaments tested directly using the FDM printer. ments could not be clearly identified. The ST and HP40 filaments were too flexible to be placed in the ATR-FTIR and PXRD were carried out to further confirm TA rig (as illustrated in Fig. 4a). The rest of the non-feedable the amorphous nature of the filaments and investigate any pos- filaments all shared a characteristic brittle fracture pattern, with sible molecular interactions between the polymers and additives. sudden discontinuation of force on the filament after reaching a Figure 3a shows the ATR-FTIR spectra of PEO and HPMCAS, peak fracture force, as seen in Fig. 4b. These filaments fractured as well as mixture filaments. The spectra of HP10 and HP20 immediately at the maximum force, showing no bending, plastic, closely resemble that of HPMCAS. The C-H stretching peaks of or elastic deformation to accommodate the increased strain. −1 semi-crystalline PEO, occurring at 2876 cm can be seen slight- Within the non-feedable filaments, Eudragit EPO, PVPVA64 ly more defined in HP30, At higher PEO concentrations the and Soluplus exhibited much sharper fracture than HPMCAS, sharp PEO bands are dominating, nearly resembling the raw HD and SP. This is evident by the longer travel distance of the PEO material, and indicating significantly increased crystallinity probe before the fracture of the HPMCAS, HD and SP filament, of PEO in these filaments. In the spectra of HP70 and HP90 the andbythe existenceofsomeminor resistance to deformationby −1 PEO peaks at 1341 and 1077 cm , corresponding to C-H and the filaments after reaching the yielding force. O-H bending, are visibly unchanged in the placebo filaments Despite seeming random at first glance, the flexibility profiles indicating no specific interactions between HPMCAS and of the feedable filaments, as seen in Fig. 5, all share a character- PEO. Figure 3b shows the ATR-FTIR spectra of PAC loaded istic bending deformation after the maximum strain bearing filaments. Across the whole spectrum, the sharp bands of crystal- point is reached. When the TA probe was returning to the start −1 line PAC are absent and a broad peak at 3321 cm ,which position, the filaments could partially recover and straighten corresponds to the N-H stretching of PAC in its amorphous state, within the rig even though they had lost some of their stiffness. can be seen in all drug-loaded filaments. In the drug containing Of the feedable filaments, filament HP10 was notable the only samples, there was no observable change in the carbonyl peak of filament to fracture during TA. However, filament HP10 did HPMCAS indicating the no significant hydrogen bonding exhibit substantial bending after reaching peak tension force interactions with the drug. The PXRD patterns of the fila- and only fractured after being bent considerably by the texture ments shown in Fig. 3c confirm the fully amorphous nature analyzer probe. Therefore, its recorded fracture pattern was of all drug-loaded filaments, as neither PEO nor PAC signals found to considerably differ from the sharp brittle fracture pat- are found in the diffraction patterns of HP10D, HP20D, and terns exhibited by non-feedable filaments, mainly HD and SP. HP30D filaments. Figure 6 shows an example of a flexibility profile combined with photographs of the filament at critical points in the profile. Filament Feedability Tests During TA testing, the filament is subjected to axial compression forces, as the TA probes continue moving towards each other, Pure polymer filaments Eudragit® EPO, HPMCAS, PVP/ the forces born by the filament continuously increase. At the VA 64, and Soluplus® were found to be too brittle and would first critical point, the filament reaches the Euler point (yield fracture inside the printing head whenever feeding was point),above whicheventhe applicationofaninfinitesimal lat- attempted. With the addition of low levels (10% w/w) of drug eral force will cause bending. This is the bending point highlight- (PAC) or plasticiser (PEG), HD and SP could still not be fed ed in Fig. 6. After this, the applied force acts to both compress the through the FDM printer. Increasing the degree of the plasti- filament and to further bend the filament at the weakened bend cization (Tween 80 20% for ST and PEO 40% for HP40) in point which leads to the complex TA profile pattern of this stage the filaments led to over-plasticization. The ST and HP40 of test as seen in Figs. 5 and 6. Notably, this characteristic feature filaments were found to be overly flexible and coiled up inside of bending and maintaining structural integrity above the Euler point was seen in all feedable filaments as shown in Fig. 5. the feeding zone and would not thread through into the Development of a Simple Mechanical Screening Method (2018) 35:151 Page 7 of 13 151 Fig. 3 ATR-FTIR spectra for HP (a)placebo filaments, (b)drug loaded filaments, and (c)PXRD diffraction patterns of the drug loaded filaments. Correlation Analysis filaments and each of the three commercial FDM printable filaments, ABS, dissolvable filament and PLA, were generated The correlations between the flexibility profiles (the normal- and listed in Table III. This correlation can be treated as the ized area under the flexibility profile) of each in-house quantification of the level of similarity between the flexible 151 Page 8 of 13 Nasereddin et al. (2018) 35:151 Table III Feedability and Correlation Coefficients of the Flexibility Profiles of In-House Pharmaceutical Filaments with the Flexibility Profiles of the Commercial Filaments Filament Feedability ABS Dissolvable filament PLA Average score Rounded mean correlation score HPMCAS N 0.38 0.64 0.18 0.40 0 Mowiflex Y 0.76 0.69 0.74 0.73 1 PEO Y 0.92 0.71 0.82 0.82 1 PVPVA64 N −0.08 0.22 −0.12 0.01 0 Soluplus N −0.56 0.40 −0.75 −0.30 0 Eudragit EPO N −0.70 0.30 −0.79 −0.40 0 EUD Y 0.94 0.65 0.88 0.82 1 HD N 0.32 0.60 0.18 0.37 0 HP10 Y 0.85 0.82 0.67 0.78 1 HP10D Y 0.96 0.53 0.94 0.81 1 HP20 Y 0.94 0.70 0.87 0.84 1 HP20D Y 0.90 0.48 0.95 0.78 1 HP30 Y 0.95 0.77 0.89 0.87 1 HP30D Y 0.91 0.40 0.94 0.75 1 HP70 Y 0.70 0.73 0.51 0.65 1 HP90 Y 0.81 0.81 0.74 0.79 1 SP N 0.49 0.63 0.28 0.47 0 Correlation matrix of the three commercial filaments ABS Dissolvable filament PLA ABS 1 –– Dissolvable Filament 0.62 1 – PLA 0.92 0.5 1 profile of the in-house pharmaceutical filaments and the com- Principal Component Analysis (PCA) mercial printable filaments. The higher the correlation score, the higher the mechanical similarity of the tested PCA is a multivariate statistical technique that, from a data filament to the commercial filaments. As seen in table containing observations describing a multitude of inter- Table III, for most of the filaments, the correlation related variables, can extract key information which is repre- scores to the 3 commercial filaments vary. This is not sented as functions of BPrincipal Components^. Similarities surprising as the correlations scores of the flexible pro- between the observations can be represented by plotting the files of the 3 commercial filaments also varies indicating variables on a map referred to as a space plot (15). PCA was there are some differences in their flexible profiles. The performed to further explore the relationship between flexi- correlation scores were further analysed by taking the mean of bility profile of the filaments and their feedability. As the cor- the three correlation scores per filament using the equation relation scores of the Dissolvable filament with the other commercial filaments are low, the dissolvable filament C þ C þ C was treated as an outlier and was not included in the ABS Dissolvable Filament PLA ð2Þ PCA. For the PCA, the normalized full force-distance curves were used the analysis. Principal Component 1 shows an eigenvalue of 10.13, Principal Component 2 where C is the correlation score with commercial fila- ment x. Overall, all the in-house filaments that passed shows an eigenvalue of 3.57, Principal Component 3 shows an eigenvalue of 1.51, while all other principal the feedability test had an average correlation score above 0.5; whereas the filaments that failed the components show an eigenvalue <1. By applying the Kaiser Rule, the first three principal components were feedability test all had a correlation score below 0.5. Furthermore, the correlation scores of plasticized fila- extracted and the component loadings are shown in Supplementary Materials Table S1. ments were higher than those of non-plasticized fila- ments indicating that plasticization improves the flexibility Figure 7a shows the rotated space plot of Principal Components 1, 2, and 3. The filaments aggregated on the plot of the filaments. Development of a Simple Mechanical Screening Method (2018) 35:151 Page 9 of 13 151 non-feedable filaments Eudragit® EPO, PVPVA 64, and Soluplus . DISCUSSION Correlation between Mechanical Properties and Feedability This study is aimed to develop a screening method to speed up the formulation development of FDM printable solid formula- tions. To achieve good printability, the FDM filaments first need to exhibit good feedability to allow the smooth and continuous delivery of the filaments to the melting zone of the printing head. As previously discussed, the feeding mechanism employed by FDM printers involves mechanical pushing of a filament held between two counter-rotating gears. In situations where the fila- ment being fed is too brittle to bear the mechanical strain gen- erated from the compression and pushing, this is likely to cause the filament to fracture, discontinuing the force that is propelling the filament forward, causing a block in the printing head. If it is too ductile it will deform when begin passed forward and again block the head. For developing pharmaceutical FDM filaments, the polymers firstly must be hot melt extrudable to form the filaments. When HMEisusedinthe fabricationoftraditional soliddosageforms (i.e. tablets), the extrudates produced often undergo particle size reduction, to produce powders or granules with a suitable parti- clesizefor pharmaceuticalprocessing (i.e. compression) (16). Unsurprisingly, brittle extrudates are more suitable in that re- Fig. 4 (a) Image of an example of a floppy filament (HP40) which was not able to be fed into the texture analysis rig for testing; (b) Texture analysis profile gard, as brittle materials require less time and energy to be milled of non-feedable filaments. or granulated as opposed to ductile/flexible materials (17). Therefore, most pharmaceutically relevant polymers that are into five clusters. The feedable filaments aggregated into three suitable for HME often yield brittle extrudates that readily frac- clusters, the first containing ABS, PLA, HP10D, HP20D, and ture, and while this makes such polymers suitable for traditional HP30D. The second contains filaments HP10, HP20, HP30, pharmaceutical applications of HME, it renders those polymers EUD, and PEO. The third cluster contains filaments unsuitable for FDM implementation. Mowiflex, HP70, and HP90. The three clusters are closely TA studies clearly show that there exists two types of non- aggregated together and can be looked at as a single macro- feedable filaments, both highly friable, either with or without any cluster. strain bearing ability. The additions of plasticizers caused an in- The fourth cluster contained filaments SP, HD, and crease in the strain bearing ability of the filaments. As an exam- HPMCAS, which are the filaments that showed some strain- ple, pure Soluplus filament exhibited no strain bearing capacity bearing ability in the TA tests but still fractured as the result of in the TA test, whereas the addition of 10% PEG (SP) shifted the compression (Fig. 4b). This cluster of slightly deformable brittle flexibility profile to the group exhibiting some strain bearing filaments is closely positioned to the macro-aggregated cluster capacity, owing to the plastisisation effect of the PEG. The in- of feedable filaments. Although the filaments in this cluster are corporation of plasticizers also decreases the glass transition, and not feedable, they exhibited some potential, such that, with improves melt flow properties of thermoplastic polymers, which formulations modification such as adding plastisisers, they can are important factors that influence FDM. Among the plasti- be possibly be tuned to become feedable. The transformation of cizers used in the formulation of pharmaceutical blends for the mechanical properties of HPMCAS upon the addition of a FDM printing are triethyl citrate, triacetin, various grades of drug (PAC) and a second polymer (PEO) is an example of such polyethylene oxides (PEG and PEO), Tween® 80, and glycerol (6,8–11). However when comparing the T tuning (Fig. 7b). The fifth cluster contains the highly brittle and data in Table II and g 151 Page 10 of 13 Nasereddin et al. (2018) 35:151 Fig. 5 Flexibility profiles of feedable filaments. (a)the commercial filaments, (b)the placebo HP filaments, (c)PAC- loaded HP filaments. Development of a Simple Mechanical Screening Method (2018) 35:151 Page 11 of 13 151 Fig. 6 Illustration of the interpreation of the force-distace of a feedable filament during TA. the feedability data in Table III, it is clear that the absolute T unduly influencing the mechanical properties. Increasing the values of the filament do not directly correlate to their feedability. PEO loading to 40% rendered the filament unfeedable Over-plasticization of filaments was observed to also cause a (HP40), which most likely is due to the significant phase separa- feeding defect; filaments HP40 and ST were found to coil inside tion of HPMCAS and PEO as indicated by the DSC in Fig. 2. the printing head when feeding was attempted. Those filaments Conversely, filaments HP70 and HP90 were found to be possess little-to-no rigidity and would readily deform when any feedable. However, the high PEO loading in comparison to force is applied, with their texture being more like that of fabrics HP10, HP20, and HP30 means that the filaments are most likely than thermoplastic polymers. This indicates that the appropriate PEO-based with the HPMCAS being the second material in the level of plasticization is vitally important. The non-feedability of matrix. This is further supported by the color difference; HP10, the overplasticized filaments is because they readily deform inside HP20, and HP30 all had the characteristic pale yellow color of the printing head making them unable to thread through the hot-melt extruded HPMCAS (18), while filaments HP70 and melting zone for deposition. This lack of rigidity sits in stark HP90 were colored identically to the PEO filament. Based on contrast to feedable filaments which are pliable enough to bend these results, it is reasonable to hypothesize that, in the case of and deform on handling, but retain their original shape when HPMCAS-PEO blends, the existence of a continuous phase (ei- force is released. ther HPMCAS or PEO) as the primary matrix former is impor- HPMCAS was selected as the platform polymer for drug tant to maintain the mechanical strength of the filaments. loading and plasticization screening because, although the fila- ment itself was not feedable, its flexibility profile clearly displayed some strain-bearing properties comparable to those of plasticized Using Flexibility Profile towards Screening Soluplus (SP). The addition of 10% PEO was found to readily transform the filament into a feedable one. Furthermore, the In terms of screening, the TA test was designed to simulate the addition of 10% PAC to the filament significantly changed its conditions inside the printing head as closely as possible, the fracture pattern from a brittle fracture to a slightly more ductile speed of compression was set to 3.15 mm/s, which matches the fracture (Fig. 7b). This can be attributed to the plasticization speed of feeding inside the printer. Tested filaments were found effect of the drug on the polymer. This is also supported by the to exist in either one of three categories; brittle filaments, string- fact that although HP10 did fracture (but was still feedable), like filaments, and pliable filaments. Brittle filaments are fila- filament HP10D did not, suggesting that the addition of 10% ments that fracture during the analysis. String-like filaments PAC further increased the strain-bearing ability of the are filaments could not be tested due to them being too flex- filament HP10. ible to maintain a vertically suspended straight beam shape Filaments HP10, HP20, HP30, HP20D, and HP30D were all and would collapse under their own weight. Pliable filaments found to be feedable, suggesting that, at least for HPMCAS, are filaments that would deform due to compression by the there is a wide margin available for plasticizer loading without texture analyzer, but recover when the force is removed. 151 Page 12 of 13 Nasereddin et al. (2018) 35:151 Fig. 7 (a) 3-dimensional rotated space plot of principal components 1, 2 and 3; (b) comparison of the flexible profiles of non-feedable, tunable and feedable HPMCAS based filaments. It should be noted that the aforementioned categories do not However, no observations were made that indicate feedability have clearly defined boundaries, but are rather like a spectrum. existing as a spectrum property of the filaments (i.e. no filaments Filaments SP, HD, and HPMCAS, despite being brittle fila- were found to be Bmore feedable than others^). Feedability of ments, did exhibit some pliability before fracturing. Inversely, the filaments is a Boolean value, being either true of false. The filament HP10 did fracture during TA, but the predominating TA data is acurve whichisthennormalizedand sorted into mechanical property it exhibited was pliability. categories by the statistical analysis. The normalisation procedure High correlation scores between the feedable in-house fila- used in this removes differences in the absolute values of force ments and the commercial filaments were observed. All feedable applied. Mowiflex is feedable but notably much stiffer than all filaments displayed correlation scores >0.50 with the commercial other tested materials, with the minimum required force for filaments, indicating the significance of the relationship between deforming Mowiflex being 120N. The higher stiffness of the the flexibility profiles of the filaments and their feedability. filament does not affect its feedability and indicates that a range Development of a Simple Mechanical Screening Method (2018) 35:151 Page 13 of 13 151 of absolute mechanical properties might lead to feedable mate- author(s) and the source, provide a link to the Creative rials provided the shape of the overall flexibility profile falls within Commons license, and indicate if changes were made. the acceptable range. To further simplify the analysis, rounding the mean correlation scores of each filament to the nearest inte- ger (values <0.5 are rounded down to 0, while values >0.5 are REFERENCES rounded to 1) produces a method (Table III) that simply sorts the filaments into feedable with a score of 1 (True) and non-feedable 1. Norman J, Madurawe RD, Moore CMV, Khan MA, Khairuzzaman A. A new chapter in pharmaceutical manufacturing: 3D-printed drug filaments with a score of 0 (False). products. Adv Drug Deliv Rev. 2017;108:39–50. PCA was used as a qualitative statistical method to sort the 2. Alhnan MA, Okwuosa TC, Sadia M, Wan KW, Ahmed W, Arafat different filaments using their flexibility profiles into feedable B. Emergence of 3D printed dosage forms: opportunities and chal- and non-feedable filaments. As seen in Fig. 7a, three clusters lenges. Pharm Res. 2016;33(8):1817–32. 3. Bakar NSA, Alkahari MR, Boejang H. Analysis on fused deposition were observed in the rotated space plot of the filament flexi- modelling performance. J Zhejiang Univ Sci A. 2010;11(12):972–7. bility profiles. The feedable and non-feedable filaments are 4. Mohamed OA, Masood SH, Bhowmik JL. Optimization of fused well separated. Interestingly a cluster containing filaments that deposition modeling process parameters: a review of current re- can be easily tuned to become feedable (referred in the Fig. 7a search and future prospects. Adv Manuf. 2015;3(1):42–53. 5. Jin YA, Li H, He Y, Fu JZ. Quantitative analysis of surface profile as ‘tunable’ filament) is also isolated. Using HPMCAS as an in fused deposition modelling. Addit Manuf. 2015;8:142–8. example (Fig. 7b), by adding different types and amounts of 6. Goyanes A, Buanz ABM, Basit AW, Gaisford S. Fused-filament 3D plasticizers, the non-feedable polymer and polymer blend printing (3DP) for fabrication of tablets. Int J Pharm. 2014;476(1):88– (HPMCAS and HD) in this cluster can be transferred into 7. Goyanes A, Buanz ABM, Hatton GB, Gaisford S, Basit AW. 3D feedable filament (HP10 and HP10D). This data demon- printing of modified-release aminosalicylate (4-ASA and 5-ASA) strates that the flexible profile obtained from TA test tablets. Eur J Pharm Biopharm. 2015;89:157–62. can be correlated to the feedability and used to predict 8. Skowyra J, Pietrzak K, Alhnan MA. Fabrication of extended- the potential of the FDM printability of the targeted release patient-tailored prednisolone tablets via fused deposition materials. modelling (FDM) 3D printing. Eur J Pharm Sci. 2015;68:11–7. 9. Pietrzak K, Isreb A, Alhnan MA. A flexible-dose dispenser for im- mediate and extended release 3D printed tablets. Eur J Pharm Biopharm. 2015;96:380–7. CONCLUSION 10. Alhijjaj M, Belton P, Qi S. An investigation into the use of polymer blends to improve the printability of and regulate drug release from pharmaceutical solid dispersions prepared via fused deposition model- Mechanical properties of the HME filaments is an important ing (FDM) 3D printing. Eur J Pharm Biopharm. 2016;108:111–25. property determining the processibility for FDM 3DP. By 11. Melocchi A, Parietti F, Maroni A, Foppoli A, Gazzaniga A, Zema measuring the flexibility, one of the most directly relevant L. Hot-melt extruded filaments based on pharmaceutical grade mechanical properties of the HME filaments, this study de- polymers for 3D printing by fused deposition modeling. Int J scribed the development of a simple method for screening the Pharm. 2016;509(1–2):255–63. 12. Huang T, Wang S, He K. Quality control for fused deposition feedability and subsequent printability of HME filament for modeling based additive manufacturing: current research and fu- FDM printing. A wide range of filaments prepared using ture trends. 2015 First Int Conf Reliab Syst Eng. 2015;56:1–6. pharmaceutical polymers and excipients were tested to vali- 13. Stokes M. 3D printing for architects with MakerBot : build state-of- date the method. The method described could accurately and the-art architecture design projects with MakerBot replicator 1, 2, or 2X. Packt Pub; 2013. reproducibly separate feedable and non-feedable filaments. 14. Boldyreva EV, Drebushchak VA, Paukov IE, Kovalevskaya YA, Furthermore, coupled with PCA, more insights were gained Drebushchak TN. DSC and adiabatic calorimetry study of the in the aspects of how plasticisation and phase separation could polymorphs of paracetamol: An old problem revisited. J Therm influence the feedability of the pharmaceutical filaments. Anal Calorim. 2004;77:607–23. 15. Williams LJ. Principal component analysis. Wiley Interdiscip Rev Comput Stat 2. 2010;2(4):433–70. ACKNOWLEDGMENTS AND DISCLOSURES 16. Maniruzzaman M, Boateng JS, Snowden MJ, Douroumis D. A review of hot-melt extrusion: process technology to pharmaceutical QIB receives strategic funding from BBSRC. The authors products. ISRN Pharm. 2012;2012(2):436763–9. 17. Iyer RM, Hegde S, DiNunzio J, Singhal D, Malick W. The impact report no conflicts of interest. of roller compaction and tablet compression on physicomechanical Open Access This article is distributed under the terms of the properties of pharmaceutical excipients. Pharm Dev Technol. 2014;19(Mcc):583–92. Creative Commons Attribution 4.0 International License 18. Lu J, Obara S, Ioannidis N, Suwardie J, Gogos C, Technical C, et al. (http://creativecommons.org/licenses/by/4.0/), which per- Understanding the processing window of hypromellose acetate succi- mits unrestricted use, distribution, and reproduction in any nate for hot-melt extrusion, part I: polymer characterization and hot- medium, provided you give appropriate credit to the original melt extrusion. Adv Polym Technol. 2016;37(1):154–66.
– Springer Journals
Published: May 31, 2018